506 Appendix C to reduce the risk of failure to an acceptable level. Blend limits for FOD and criteria for removal or repair should be based on the FOD analysis required in the preceding paragraph. Material allowables for material which has been repaired, such as by blending or welding, should be adjusted as necessary to account for any degradation of the fatigue limit due to the blend or repair operation. The effect of any redistribution of internal stress, whether induced by FOD or by prior processing, should be considered in the development of safe blending limits. Other Damage States Material allowables should be established based on any other damage state as described in section A.4.6 under the “High Cycle Fatigue” heading. Analytical studies have shown that not every part location will be limited by a crack growing to a calculated critical stress intensity equal to the material’s fracture toughness. Some part locations will in fact be life-limited by cracks growing to a predicted vibratory threshold K th HCF, where K maxallowable LCF =K th HCF/1 −R and R = steady − vibratory steady + vibratory steady =maximum operating stress neglecting vibratory stress vibratory =1/2(peak to peak vibratory stress) and K th HCF =fRtemp Overspeed residual strength requirements need not be considered for those part locations limited by cracks reaching a calculated vibratory threshold. One overspeed cycle occurring at a crack size equal to the vibratory threshold creates less damage (change in crack size) than additional LCF–HCF crack growth from the vibratory threshold to a maximum stress intensity K CRIT defined by the material fracture toughness. For those locations not limited by vibratory stress concerns, the part’s maximum allowable crack size should be limited to a size that will survive the maximum design stress that occurs on the last cycle of the calculated safety limit. REQUIREMENT GUIDANCE (A.4.8.5) The flaw growth interval is also known as the safety limit. It is recommended that the flaw growth intervals be twice the inspection intervals specified in A.4.8.4. Flaw growth inter- val margins, other than two, can be used when individual assessments of variables (i.e., initial flaw size, da/dN , kiC, etc.) that affect flaw growth can be made (e.g., to account Appendix C 507 for observed scatter in crack growth during testing). The following should be considered in treatment of variables which can affect the calculation of the flaw growth interval: (a) No life credit should be taken of any beneficial effects of residual stresses or surface treatments such as shot peening or coatings, except when the following conditions are met: (1) The beneficial effect of residual stresses or a surface treatment can be verified through analysis and testing for the duration of the service interval and under the actual anticipated usage and maintenance conditions such as polishing for eddy current inspection. (2) Quality control procedures for application and in-service integrity are demon- strated to a satisfactory level of reliability. These beneficial effects should be verified and the extent of life “crediting” should be approved by the Procuring Activity. (b) Damage in a primary structure may result in load increases in the secondary structure. The analysis of such secondary structures should account for this. (c) Continuing damage should be assumed at critical locations where the initial damage assumption does not result in failure of the part (e.g., the case of a free surface at a bolthole). The following assumptions of initial damage and location should be consid- ered with the limiting condition used to establish safety limits and inspection intervals: (1) When the primary crack and subsequent growth terminates prior to component failure, an initial flaw equal to or greater than that which is demonstrated to be inspectable in 4.8.3 should be assumed to exist at the opposite side of the feature after the primary crack has terminated. The stress gradient assumed at the opposite location should be based on the boundary conditions that exist when crack growth has terminated at the primary location. The safety limit for this condition should be the sum of the crack growth at the primary location and at the opposite location. (2) Growth of an assumed initial flaw at the location opposite the primary location should be verified as an initial condition. (d) The effects of vibratory stress on unstable crack growth should be accounted for when the safety limit is established. Threshold crack size should be established at each individual sustained power condition (Idle, Cruise, Intermediate) using the appropriate values of steady stress and vibratory stress. The largest threshold crack size should be used as a limiting value in calculation of the safety limit if it is 508 Appendix C less than the critical crack size associated with the material fracture toughness. An analytical approach to defining the effects of vibratory stress is based on a maximum stress intensity allowable, K maxallowable LCF (as a function of stress ratio, R), which is predicted from appropriate material HCF vibratory threshold K th HCF (as a function of stress ratio, R) properties at steady-state operating conditions. This relationship is as follows: AssumeK maxallowable LCF =K th HCF/1 −R where R = steady − vibratory steady + vibratory steady = max operating stress neglecting vibratory stress vibratory =1/2peak to peak vibratory stress and K th HCF =fRtemp K th HCF versus R-ratio material property curves used in this verification at various temperatures should be developed during material characterization, as necessary. (e) The K allowable for threshold crack growth rate should be based on the crack length under consideration; the maximum allowable crack extension which will not produce failure, instability, or measurable change in dynamic response characteristics; and number of HCF cycles between inspections for damage-tolerant components, or full fatigue life (per section A.4.13.3) for durability-limited components. It should be demonstrated, through analysis or testing, that the limit in the number of allowable HCF cycles should not be exceeded as specified above for the values of da/dN th and K th chosen for the definition of the threshold condition. For conditions where da/dN versus K is not known (for example, in the small crack regime or where a crack has not formed to a measurable or deterministic size), then the threshold condition should be based on a fatigue limit (as specified in section A.4.13.3) for the required number of cycles. In general, the threshold crack growth rate condition, da/dN th , and the fatigue limit that corresponds to both the appropriate number of HCF cycles required by the design condition and to the maximum allowable crack size or damage state should be met. (f) Galling/fretting limits (i.e., permissible depth of surface damage) for all contact surfaces should be defined based on K maxallowable LCF (as a function of stress ratio, R) or stress analysis that demonstrates that the fatigue limit is not exceeded for the specific part under the applied steady and vibratory stresses. (g) Calculation of hold time crack growth. Assessments of material/design acceptability do not typically account for the flaw growth under constant load conditions. The duty cycle is compressed such that only the cyclic content is preserved. If the constant load duty cycle content is neglected, the rate of growth of cracks can Appendix C 509 be underestimated. Early assessments of material/design candidates should include the effects of hold time crack growth under representative load and temperature conditions. Initial screening and subsequent re-screening as the duty cycle matures should be performed to pre-empt service failures. REQUIREMENT LESSONS LEARNED (A.4.8.5) Since average fracture properties have been used in analysis, parts made from materials with scatter factors greater than two have failed prior to their inspection interval. Thus, for materials with large scatter factors (i.e., greater than two), factors of safety greater than two, on residual life, should be considered. In addition, the USAF has experienced several disk post (lug) failures attributed to high stress gradients arising at the Edge of Contact (EoC) between the blade and disk. Conventional lifing assessment practices did not address the susceptibility of the material/design to prevent the cracks from propagating under hold time crack growth conditions. A.5.8.5. Flaw growth. The requirements of 4.8.5 should be verified by analyses and tests. VERIFICATION RATIONALE (A.5.8.5) Verification of flaw growth is necessary to ensure initial flaws will not grow to critical size and cause failure due to the application of the required residual strength load. VERIFICATION GUIDANCE (A.5.8.5) The test should be conducted in accordance with A.4.8. Analyses should demonstrate that the assumed initial flaws will not grow to critical size for the usage, environment, and required damage tolerance operational period. The analyses should account for repeated and sustained stresses, environments, and temper- atures, and should include the effects of load interactions. Analysis methods should be verified by test, utilizing engine and rig testing. VERIFICATION LESSONS LEARNED (A.5.8.5) None. 510 Appendix C A.4.13.3. High cycle fatigue. The probability of failure due to HCF for any component within or mounted to the engine should be below 1 ×10 −7 per EFH on a per-stage basis, provided the system-level safety requirements are met. REQUIREMENT RATIONALE (A.4.13.3) High cycle fatigue has been a major safety and maintenance problem in jet engines. Proper attention is required to minimize these failures. A requirement based upon probability of failure is most consistent with HCF experience, given the inherent variability of the many factors involved. REQUIREMENT GUIDANCE (A.4.13.3) General guidance for HCF design is provided here. Specific guidance for HCF design is provided in A.4.13.3.1 through A.4.13.3.3. Variations in the endurance capability of the material, determination of dynamic stresses, determination of steady-state stresses (or pseudo-steady-state stresses), the sequence in which combinations of stresses occur, and other factors all affect the determi- nation of HCF probability of failure. Indeed, this observation has led to the establishment of the probability of failure requirement. The HCF design will be a function of proba- bilistic design margins on frequency, predictions on the variation on alternating stress, and the current threshold-based approach for material capability. The probabilistic design margin on frequency will give the probability of resonance for a mode at a given operational condition and should be compared to the system-level reliability requirement. It is conservatively assumed the probability of resonance equals the probability of failure for modes determined to have significant modal excitation (i.e., low order modes and adjacent upstream engine order excitations). Probabilistic design margins on frequency should be computed on the basis of steady-state operation at low order crossings and all known drivers within two stages either upstream or downstream of the subject component. Variations in resonance conditions should be accounted for through probabilistic analysis or appropriate engine test data distributions for the system design condition. Probabilistic design margins lead to a Probabilistic Campbell Diagram, which follows. Deterministic design margins of 10% may be used for preliminary design or when there is insufficient confidence in probabilistic solutions. In the event an insufficient probabilistic design margin on frequency exists to meet the probability of failure requirement, the next level of probabilistic design analysis is required to predict the variation in resonant stress response. Such probabilistic design Appendix C 511 analysis includes physical models of forced response, damping, and mistuning, along with the appropriate probability models for each random variable and correlations between those variables, as appropriate. It should be shown that at the resonant condition that was not avoided with the probabilistic design margin on frequency, the vibratory stress distribution is at a level that meets system reliability requirements when the stress is integrated with material capacity to predict failure. A deterministic design margin may be used when there is insufficient confidence in the probabilistic solution. Refer to the Figure C.1 below for margins on the material capacity and the current threshold-based approach, and to sectionA.4.6fordefinitionofmargin-relatedterminology. Root mean square component vibratory stresses should not be within 60% of the min- imum material endurance capability (i.e., limited to a maximum of 40% of the minimum material endurance capability). If instantaneous peak component vibratory stresses are used, they should not be within 40% of the material endurance capability (i.e., limited to a maximum of 60% of the minimum material allowance capability). Minimum material endurance capability is defined by S/N fatigue tests of a statistically significant population of fatigue specimens at various combinations of steady and alternating loads, or R-ratios, representing various levels of local component steady-state or pseudo-steady-state stress. All engine parts should have a minimum HCF life of 10 9 cycles. This number is based on the observation that an endurance limit does not exist for most materials. If it can be shown through analysis or test that a given part will not experience 10 9 cycles during its design life, a number lower than 10 9 may be used. Such a condition may be established through analysis of vibrating frequencies and probabilities of a given part being subjected to steady-state or transient vibrations. It should be shown that the total time of exposure to any frequency and amplitude is less than 10 9 cycles or that the amplitude is less than the material allowable at 10 9 cycles. An alternate approach is to use a life of 10 9 cycles based on data obtained at shorter lives, but not less than 10 7 cycles, and a demonstrated valid method to extrapolate to 10 9 cycles to establish an endurance Vibratory stress Steady stress Margin –3 σ material endurance capability Figure C.1. Material capacity margins. 512 Appendix C limit. Cycles which have vibratory stress amplitudes less than the endurance limit at 10 9 cycles can be considered to have no detrimental effect on pristine material and can be ignored in damage accumulation evaluation, provided no other damage is present (see section A.4.6). For airfoils with FOD/DOD damage tolerance requirements of K f =3, for instantaneous peak values of component vibratory stress, the alternating stress should be limited to 40% (for RMS values −27%) of the minimum unnotched HCF material allowable or 100% of the K f = 3 minimum notched HCF material allowable. Further guidance to establish damaged material capability is provided in section A.5.6. One-hundred percent of the Goodman allowable may be used for components with surface enhancements, such as laser shock peening (LSP). When this is done, a thresh- old analysis and a B.1 life must be established based on data measurements from aeromechanical testing. The minimum life must be based on 1 per the anticipated number of engines in the fleet or no less than 1 per 1000 engines, as a minimum. In other words, there can be no more than 1 engine with 1 blade having an HCF life occurring at one times 1× the number of blades in the stage for the entire fleet or 1000 engines, whichever is more. REQUIREMENT LESSONS LEARNED (A.4.13.3) Complications exist with the concept of specifying all parts be designed to some discrete specified endurance limit. Some of these are: (a) Prior stressing at a higher stress can cause a lowering of the endurance limit. (b) Stress cycling at gradually increased cyclic stress can result in an increased endurance limit (this is known as “coaxing”). (c) Interactions between LCF and HCF can result in either increased or decreased lives depending upon the magnitude of the loads, the order of the loading, and the material. This is referred to as “load sequencing.” This phenomenon is evidence that HCF margin determination cannot be defined accurately without consideration of the overall stress-state over time. (d) Installation, handling, and environmental sensitivities can result in significantly higher steady-state and vibratory stresses which will reduce or even have negative margins for HCF capability. Such an example would be external parts which may be sensitive to all of the above. Realistic levels of stress due to these sensitivities should be included when HCF capability is assessed. Because these complications exist, future efforts should be aimed at components designed to the HCF probability of failure requirement. Future efforts should also strive to integrate Appendix C 513 HCF-related damage with other forms of damage (e.g. LCF, creep, etc.) by full consid- eration of the load sequence and the response of the material to that load sequence. A.5.13.3 High cycle fatigue. Verification of the engine’s ability to withstand HCF should be through analysis and test. Probabilistic design margins and predictions should be validated with bench, rig, and engine test experience in addition to statistical comparisons to operating fleet databases. Assurance is to be provided by verifying that the probability levels for each contributing random variable are within the experimental data range for that variable. VERIFICATION RATIONALE (A.5.13.3) High cycle fatigue is a very complex problem. A great deal of testing and analysis is necessary to avoid HCF problems in the field. VERIFICATION GUIDANCE (A.5.13.3) When a validated design system is not in place, a method to extrapolate empirical data should be used to demonstrate compliance with 4.13.3 (e.g., integrally bladed rotors [IBR’s] validated with F100 data). Specific guidance for HCF is provided in sections A.5.13.1 through A.5.13.3. Because of the complex nature of HCF, maximum effort should be expended to insure that all information gathered at all steps of the design and verification process are leveraged. This approach is referred to as “an holistic approach.” Models developed during the earliest stages of the design process can be used to assess the sensitivity of mode shapes and frequencies to geometric variations and variations in boundary conditions or operational conditions (referred to as “influence parameters”). Efforts should be aimed at maintenance of consistent modal characteristics over the range of geometric tolerances and influence parameters the part may experience. Design models should be carried forward to the verification process. The verification process should include accepted practices for validation of the models such as the modal assurance criteria (MAC) and others. (For more information on MAC, see Shock and Vibration Handbook, Cyril M. Harris, 4th edition, McGraw Hill, New York, 1995 and its cited references.) Once models are verified, they should be used to establish opti- mal instrumentation locations for subsequent tests. Criteria to define optimum locations include mode sensitivity (the ability of the sensor to detect maximum mode amplitude), mode identification (the ability to distinguish between modes of similar frequency), and 514 Appendix C other physical criteria such as lead routing, proximity of sensors, and the like. Varia- tions in part geometry can be addressed using sensitivity-based approaches applied to a nominal-geometry model. After verification, models can be used to define vibratory and steady stress fields for all component locations and at engine operating conditions. These normalized stress fields can be scaled to results derived during experimental test to establish the stress time history (load sequence). Should the component fail to meet HCF design requirements, the verified models can be used to direct redesign efforts. Verification tests in a lab environment (shaker table), in rigs, and in full-up engines generate large volumes of strain data. All these data should be archived to establish a database of responses that can be used to assess variabilities and be used in the validation of probabilistic predictions. Further, examination of all relevant data is useful to define the robustness of a given design over the range of variables tested. Part-to-part variability and the variation of responses to influence parameters—like local pressure, temperature, or flow angularity—can be assessed using these data to define statistically what the maximum expected response may be during operational deployment. One-hundred percent of the Goodman allowable may be used for components with surface enhancements, such as LSP. When this is done, a threshold analysis and a B.1 life must be established based on data measurements from aeromechanical testing. The minimum life must be based on 1 per the anticipated number of engines in the fleet or no less than 1 per 1000 engines as a minimum. In other words, there can be no more than 1 engine with 1 blade having an HCF life occurring at one times 1× the number of blades in the stage for the entire fleet or 1000 engines, whichever is more. VERIFICATION LESSONS LEARNED (A.5.13.3) Historical approaches to HCF verification have relied almost entirely on experimental results. Design models were generally used only to establish whether the design should be fabricated and to define the steady stress field at some assumed critical operating condition. Once fabricated, the locations of maximum vibratory stress, determination of critical locations, and HCF margin were done almost exclusively using experimental methods. Newer designs have features that have made an entirely experimental verifica- tion approach less reliable. Low aspect ratio blades tend to have a large number of modes closely spaced in frequency within the operating range of the engine. This makes mode avoidance difficult. It also makes identification of modes using response frequency alone more difficult. Mode shapes and frequencies of low aspect ratio blades tend to be more sensitive to small variations in geometry. This leads to greater variability in blade-to-blade stresses and HCF margin—a non-robust characteristic. Appendix C 515 Newer designs have also tended to employ integrally bladed rotors; i.e., conventional dovetail or firtree attachments have been eliminated. This has led to reductions in damping for rotor systems and increased complexity system dynamics. Such newer designs also tend to admit a higher degree of disk-blade coupling than earlier, heavier designs. Such coupling is known to increase the probability of HCF failure events with significant potential for uncontained failure modes. All of these factors have necessitated an evolution in the verification process to integrate analytical techniques and experimental approaches. It also shows the importance of using probabilistic approaches to predict the variation response that need to be validated with test and historical databases. A.4.13.3.2. Component vibrations. Engine components should be free of detrimental resonance at all speeds in the operating range. This can be accomplished by intentionally designing modes out of the engine operating speed range or by providing sufficient damping, a probabilistic design margin on frequency and a probabilistic prediction of vibratory stress with respect to steady-state operating speeds, or excitation control to ensure that modes which remain in the running range do not respond detrimentally. A detrimental response is one that exceeds criteria outlined in A.4.13.3. REQUIREMENT RATIONALE (A.4.13.3.2) Resonance conditions should be avoided or controlled so that amplified response and structural failure does not occur. Margin, as defined in A.4.13.3, is required due to fre- quency variations that can occur among a population of engines; or because of changes in operational conditions, deterioration, distortion, or combinations thereof; and other impor- tant random design variables. Experience has shown that engine structural components operating under combined steady and vibratory stress conditions should be designed to ensure resistance to HCF cracking. REQUIREMENT GUIDANCE (A.4.13.3.2) Resonances in the engine-operating speed range should not occur at steady-state operating speeds such as, but not limited to, idle, carrier approach, hover, cruise, or maximum. Sufficient frequency margin, as defined in the A.4.13.3, and established during normal operation of a nominal engine at sea-level conditions, should be provided to insure resonances at steady-state operating conditions do not occur elsewhere in the flight . capability). Minimum material endurance capability is defined by S/N fatigue tests of a statistically significant population of fatigue specimens at various combinations of steady and alternating. tests in a lab environment (shaker table), in rigs, and in full-up engines generate large volumes of strain data. All these data should be archived to establish a database of responses that can be. range of variables tested. Part- to -part variability and the variation of responses to influence parameters—like local pressure, temperature, or flow angularity—can be assessed using these data to